Introduction
Artificial intelligence (“AI”) is rapidly becoming part of the modern workplace. Employees are using AI to draft correspondence, prepare reports, conduct research and perform administrative tasks, while employers are increasingly using AI in recruitment, performance management, monitoring and decision-making.
The question is no longer whether AI will affect the employment relationship, but how existing labour-law principles will apply to its use.
Can an Employee Be Disciplined for Using AI?
There is nothing inherently wrong with an employee using AI as a workplace tool. Problems may arise, however, where its use is unauthorised, contrary to company policy, results in inaccurate work, or involves the disclosure of confidential or personal information.
An employee cannot simply argue that “AI made the mistake”. Where the employee remains responsible for the work, the ordinary duties of care, competence and compliance with lawful and reasonable instructions continue to apply.
Employers should therefore have clear policies regulating the acceptable use of AI.
Who Is Responsible for an AI-Generated Decision?
This is potentially more significant.
Employers may increasingly use AI to screen job applicants, assess performance, monitor productivity or identify employees considered to be underperforming.
But an employer cannot avoid responsibility for an unfair decision simply because it was generated or recommended by an algorithm.
AI systems can reflect inaccurate, incomplete or biased data. The fact that a decision is generated by technology does not necessarily make it objective or fair.
The fundamental principles of South African labour law remain applicable: fairness, equality, non-discrimination, procedural fairness and the requirement that employment decisions be based on relevant and reliable information.
AI and Disciplinary Proceedings
AI may also become part of the evidence presented in disciplinary proceedings. Employers may use technology to analyse attendance records, communications, productivity information or other workplace data.
However, an AI-generated conclusion should not simply be accepted as fact. The underlying evidence should be capable of being tested and explained.
A disciplinary chairperson must ultimately determine whether the misconduct has been established on the evidence presented. AI may assist with identifying or analysing evidence, but it should not replace the decision-maker.
AI and Retrenchments
AI may also result in the automation of functions previously performed by employees. Where this results in a genuine reduction in the need for employees, the employer may potentially have operational-requirements grounds for restructuring.
However, the introduction of AI does not remove the employer’s obligations under the Labour Relations Act. Appropriate consultation and consideration of alternatives remain essential.
What Should Employers Do?
Employers should consider adopting an AI policy dealing with:
- permitted and prohibited uses of AI;
- confidentiality and personal information;
- verification of AI-generated information;
- employee responsibility for AI-assisted work;
- AI in recruitment and performance management;
- monitoring and surveillance;
- human oversight of AI-generated decisions; and
- disciplinary consequences for misuse.
The International Labour Organization has identified algorithmic management and AI-assisted human-resource decision-making as significant emerging workplace issues, while warning that poorly designed systems and biased or incomplete data can produce unfair outcomes.
Conclusion
AI may transform the workplace, but it does not replace the employer’s legal responsibilities.
The principle is simple:
AI can assist in making an employment decision. It should not be allowed to make the decision without meaningful human oversight.
South African employers would therefore be well advised to establish clear rules now, before an AI-related dispute reaches the CCMA.
